An Integrated Dataset for Dynamic Avoidance Rules and Safety Separation in Low-Altitude Aircraft

Low-Altitude Air Mobility (LAM) encompasses heterogeneous aircraft, such as eVTOLs and other low-altitude vehicles. The safe operation of this airspace relies heavily on robust Conflict Detection and Resolution (CD&R) mechanisms and engineering-feasible separation standards. Addressing the limitations of traditional static separation standards based on fixed distance or time, this paper proposes a dynamic safety separation model for low-altitude aircraft and introduces an open-source dataset named LowAltAvoid. This dataset contains both simulated and measured data from over 1,000 flight scenarios, detailing eight flight scenarios under 5 types of complex meteorological disturbances and navigation errors. It not only covers the time-varying safety separation envelopes determined by kinematic constraints, navigation uncertainty, control latency, and environmental disturbances but also realizes the integrated recording of these envelopes coupled with dynamic avoidance rules, including priority, maneuver logic, response timing, and maneuver types. LowAltAvoid encompasses diverse elements such as flight corridors, gridded airspace, vertical tubes, and mixed traffic. It systematically includes multi-dimensional data on conflicts between low-altitude aircraft as well as between low-altitude and manned aircraft, specifically providing synchronized state trajectories, conflict detection results, Closest Point of Approach (CPA) metrics, recommended avoidance actions, and corresponding dynamic safety thresholds. Consequently, the LowAltAvoid dataset serves as a high-confidence benchmark for validating learning-based collision avoidance algorithms, significantly enhancing the engineering applicability of low-altitude safety decision-making technologies.